PR #14317 ggml-cpu: enable IBM NNPA Vector Intrinsics

ggml-cpu: add nnpa compile flag

5801806f

ggml-cpu: add fp16->fp32 nnpa first

45a4cf65

ggml-cpu: add fp32->fp16

ebf9f34a

ggml-cpu: better variable names

ffe29645

docs: update s390x docs

0394a006

ggml-cpu: add debugging prints to see if dlf16 is correct

48b820d0

ggml-cpu: fix print vs printf

d9cc63a9

ggml-cpu: fix float placeholder

94f10ca1

ggml-cpu: ensure fp16 and fp32 load and stores are called

8f3a5af6

ggml-cpu: fp16 load ensured to hit

575ea9f6

ggml-cpu: remove sigint from fp16 store

9330454c

ggml-cpu: activate nnpa for ggml_cpu_fp16_to_fp32

ebc1d19f

ggml-cpu: nnpa activate ggml_cpu_fp16_to_fp32 for 8 elements

6a25fd85

ggml-cpu: nnpa switch to vec_xst test

f9f6c7e8

ggml-cpu: switch to vec_xst for 4 element loops also

6d507bbe

ggml-cpu: rework noop

8312adc9

ggml-cpu: remove noop, general code cleanup

27b4c3f3

ggml-cpu: clarify variable naming

e0f8fb93

ggml-cpu: activate nnpa for ggml_cpu_fp32_to_fp16

bb9345ca

ggml-cpu: add breakpoint for debugging

5424d9e7

ggml-cpu: test fix for conversion failure

4f017d71

ggml-cpu: disable fp32->fp16 nnpa conversions for now

27131e5f

ggml-cpu: switch to elif macro

946c78eb

ggml-cpu: reattempt fp32->fp16

433d5874

ggml-cpu: fix typo

54811fc1

ggml-cpu: reattempt fp32->fp16

e12e9fe7

ggml-cpu: fix compiler types

7413dabc

ggml-cpu: change to typedef vector types

373fa28e

ggml-cpu: add 4 element loops for fp32->fp16

4621a23c

ggml-cpu: clarified vector naming

987d1690

ggml-cpu: bring back fp32->fp16 store nnpa

8ef51b90

ggml-cpu: activate nnpa fp32->fp16 or fp16->fp32 compute

f1b1d98e

ggml-cpu: add nnpa macro check in ggml-impl

1547ea23

ggml-cpu: add missing __func__

0e571dd3

ggml-cpu: diagnose why __NNPA__ macro is not being defined

4ad6efa3

ggml-cpu: import vecintrin.h to fix compiler errors

81298380

ggml-cpu: update macro tests

e7910fc9

ggml-cpu: move s390x typedef to own header file

157f856c

Revert "ggml-cpu: move s390x typedef to own header file"

48df9770

ggml-cpu: switch to importing ggml-cpu-impl instead

3004a79f

ggml-cpu: fix macro declaration

1cacdd9a

ggml-cpu: test more macros

fadc1387

ggml-cpu: add debug prints

ed76ff6e

ggml-cpu: bruteforce macro definitions

84593387

ggml-cpu: move macro definitions

72c91436

ggml-cpu: add ggml-impl.h to cmakelists

a91c3ab6

ggml-cpu: switch to private macros

ba3513e4

ggml-cpu: move s390x typedef to own header file

18d79e1a

ggml-cpu: move things around

781c2637

ggml-cpu: bring back compile macros

263b820b

ggml-cpu: switch to quotes for import

04a395ea

ggml-cpu: add compiler error macro

c8b3b895

ggml-cpu: add s390x detection in ggml-src

ebb8489a

ggml-cpu: bring back compile definitions

3ec0bdc1

ggml-cpu: undo cmakelists work

e43dc82a

Revert "ggml-cpu: move s390x typedef to own header file"

5c9b0835

ggml-cpu: remove typedefs.h

1b4dbf47

ggml-cpu: remove typedef from cmakelists

46227c61

ggml-cpu: add ggml-impl.h future notes

72965ea8

ggml-cpu: add todo comment for future reference

07de57c6

github-actions added documentation

github-actions added ggml

ggml-cpu: clarify naming of dlf16

489cdf44

ggml-cpu: remove unnecessary target compile definitions

5004e439

ggml-cpu: move nnpa fp16->fp32 and fp32->fp16 to simd-mappings

5834dee1

slaren commented on 2025-06-23

ggml: refactor fp32->fp16 and fp16->fp32 simd to ggml-cpu

bd288e8f

docs: update broken huggingface link for s390x

4d136cb6

ggml-cpu: fix duplicate func names during compile

fbb73345

Revert "ggml-cpu: fix duplicate func names during compile"

e73413bb

Revert "ggml: refactor fp32->fp16 and fp16->fp32 simd to ggml-cpu"

8a5e011c

ggml: refactor fp16<->fp32 simd to ggml-cpu

17b032fa

ggml-cpu: fix missing simd-mappings.h import in quants.c

0367b803

ggml-cpu: fix missing simd-mappings.h within repack

e615f73b

slaren commented on 2025-06-24

ggml-cpu: fix amx mmq missing simd-mappings.h

3c055a42

ggml-cpu: attempt at fixing loongarch failing build

e4666f93

ggml-cpu: move nnpa together with other fp16<->fp32 simd

e4a7f84d

ggml-cpu: fix wrong refactor of ggml-base

1e6ebb2b

ggml: remove dependency on ggml-cpu from ggml-base

64568ffb

taronaeo requested a review from

slaren 336 days ago

slaren commented on 2025-06-24

ggml-cpu: rename all fp16<->fp32 macros to prefix with ggml_cpu

a02b360f

ggml-cpu: remove mistaken fallback macro

1b23fec0

slaren requested a review from

copilot-pull-request-reviewer 336 days ago

copilot-pull-request-reviewer commented on 2025-06-24

slaren commented on 2025-06-24

ggml: move ggml_table_f32_f16 to ggml-cpu

9e40d984

ggml-cpu: move ggml_table_f32_f16 back to ggml-base due to ci failures

32a35335

Revert "ggml-cpu: move ggml_table_f32_f16 back to ggml-base due to ci…

827fce9c

Revert "ggml: move ggml_table_f32_f16 to ggml-cpu"

5be39c11

ggml: move ggml_table_f32_f16 to ggml-cpu

59b48e48

ggml: move ggml_table_f32_f16 to ggml-cpu.c

6cebee25

slaren commented on 2025-06-25

ggml-cpu: extern c ggml_table_f32_f16 + chore docs

5f2a09a8

ggml-cpu: dedup ggml_table_f32_f16 from simd-mappings.h

f71b21d2

Revert "ggml-cpu: dedup ggml_table_f32_f16 from simd-mappings.h"

176e1db6

ggml-cpu: bring back ggml_table_f32_f16

2dce1191

Revert "ggml-cpu: bring back ggml_table_f32_f16"

bb35ea67

fix ggml time initialization

8efdc0b5

fix f32_f16 table init

4ce16fa0

slaren commented on 2025-06-25

remove extra line

97620ac1

slaren approved these changes on 2025-06-25

slaren merged 60ef23d6 into master 335 days ago

llama.cpp
ggml-cpu: enable IBM NNPA Vector Intrinsics
#14317

Merged

ggml-cpu: enable IBM NNPA Vector Intrinsics #14317

llama.cpp ggml-cpu: enable IBM NNPA Vector Intrinsics #14317 Merged

ggml-cpu: enable IBM NNPA Vector Intrinsics #14317

llama.cpp
ggml-cpu: enable IBM NNPA Vector Intrinsics
#14317

Merged